Release CPU stress test


notaz

Certified Guru
Joined
Aug 23, 2005
Messages
4,913
Location
Lithuania
Website
notaz.gp2x.de
This is a NEON-based stress test that tries to automatically find your highest stable overclock while NEON is active. As it's a stress test, the value it reports will be lower than your usual stable overclock (much lower than the one that can be used on programs without NEON). This is somewhat similar to SSE-based stress tests like prime95 on PC that overclocking people use.


Note that you should run this for at least a few hours, preferably overnight (you should plug it to AC in that case as power drain is much higher than usual too). Also note that it will climb up at first, then hit an error and start climbing down, as the short pauses to switch to higher clock are enough to give CPU some rest and errors to be avoided at initial phase.


Hopefully this will help people find optimal "safe" overclock to use with PCSX, M-HT's filters and such. Get it from repo, as usual:


http://repo.openpand...lock_test.notaz
 
Last edited by a moderator:
Running.


Indeed, i run my unit @ 1GHz, but i'm now stepping down @ 956MHz with errors TT !
 
mine maxes out at 987 :D


EDIT: hmm... down to 980... still rather high for a 256Mb panda
 
Last edited by a moderator:
Very useful tool! Would be nice if everyone could put their max overclock speed in this thread, together with the machine type (orig, rebirth, 1GHz). I would be interested in getting some statistics on this, so we can say things like "everyone can get 750MHz, and at least 8 in 10 can get 900MHz" with some accuracy.
 
Does this save its output anywhere? The terminal buffer isn't very big :)


The only visible info is starting at 1000MHz, and now stepping down - currently at 969MHz and stepping down. I usually run my panda at 950MHz without issues so this will be interesting! It's been running overnight.


D.
 
I think what it does is first go up with a single test each time, so it does not discover all errors. Then when it finds the first error, it goes down, and it keeps testing at each clockspeed until it finds an error. So this does not terminate, but if at a given clockspeed, the stresstest does not find any error within, say, one hour, it's probably safe to consider that clockspeed to be OK.


So ZXDunny: that probably means 969MHz is your safe top speed, if it's still at that speed after one night.


PokeParadox: are you sure your input is needed? In the going-up phase, it doesn't produce newlines, it overwrites the line each time.
 
PokeParadox: are you sure your input is needed? In the going-up phase, it doesn't produce newlines, it overwrites the line each time.

I know what he's talking about: if you have warnings turned on for overclocking, it shows that warning in the terminal and asks for y/n. I just disabled the warning.
 
PokeParadox: are you sure your input is needed? In the going-up phase, it doesn't produce newlines, it overwrites the line each time.

I know what he's talking about: if you have warnings turned on for overclocking, it shows that warning in the terminal and asks for y/n. I just disabled the warning.
Aha... that makes sense. Thanks! :)
 
errors detected at 979MHz, stepping down - Pandora


errors detected at 850MHz, stepping down - Wat R U doin


errors detected at 631MHz, stepping down - Pandora


errors detected at 425MHz, stepping down - Stahp!
 
Ouch, does this mean Fahrstuhl's cpu is broken?


More data points please! And please mention the model type!


So far we have:


- original (CC) Pandora: 778 MHz (onpon4), 980 MHz (Takedown; was this the final number?)


- Rebirth (GC) Pandora: 999 MHz (_wb_)


- unknown Pandora: 969 MHz (ZXDunny), 956 MHz (apersonthinks), 956 MHz (Linux-SWAT; was this the final number?)


- no 1GHz Pandora data points yet
 
My Results so far.Rebirth (GC) Pandora: 994MHz 985MHz Edit: didn't realize it was still going.. after some stepping down it's now running at 985MHz for the last 20 minutes without errors.


First Batch (CC) Pandora: TBD - will test when I get home.
 
Last edited by a moderator:
On my GHZ Pandora, stable at 1220 MHZ / OPP4 B)


(still testing)
 
Last edited by a moderator:
Back
Top